AI Paper Atlas

Research map / Large language models / Reasoning and chain of thought

TIV: Thought Injection via Vectors for Efficient Reasoning in Large Reasoning Models

Yi Cao, Weijie Shi, Wei-Jie Xu, Yucheng Shen, Yue Cui, Hanghui Guo, Shimin Di, Ziyi Liu, Jiaming Li, Alexander Zhou, Jia Zhu, Jiajie Xu

AAAI 2026 · Large language models › Reasoning and chain of thought · approach: Efficient and adaptive thinking · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes TIV, a framework that compresses token-level reasoning traces into thought vectors that are injected to make large reasoning models reason more efficiently and reduce overthinking.

Problem

Large reasoning models often overthink, generating excessive tokens that add little to final accuracy while increasing inference cost.

Approach

Compresses token-level reasoning into vector representations ("thought vectors") and injects them back into the model to guide reasoning with fewer tokens, rather than relying on explicit token-level chains.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add vector injection to the compression stage (ablation)TIV (R1-1.5B) · AIME 2024AES: 0.03 -> 0.78Improved
add vector injection to the compression stage (ablation)TIV (R1-1.5B) · GSM8KAES: 0.41 -> 0.65Improved
add length penalty to the reward (ablation)TIV (R1-1.5B) · MATH 500avg tokens: 2811.1 -> 1283.3Improved
add length penalty to the reward (ablation)TIV (R1-1.5B) · AIME 2024AES: 0.63 -> 0.78 AES, accuracy 30.8 unchanged, length 9021.0 -> 6605.4Mixed
add Stage I reasoning calibration (ablation)TIV (R1-1.5B) · AIME 2024accuracy (%): 24.2 -> 30.8Improved
inject vectors after MLP as well as MHA (ablation)TIV (R1-1.5B) · AIME 2024accuracy (%): 30.8 -> 32.5Improved
increase number of injected vectors beyond one (ablation)TIV · training reward and lengthrewardNo clear effect
increase rollouts per prompt beyond 5 (ablation)TIV · training reward and lengthrewardNo clear effect
apply TIV to R1-7B reasoningDeepSeek-R1-Distill-Qwen-7B · AIME 2024accuracy (%): 52.9 -> 48.5Hurt
apply TIV to R1-7B reasoningDeepSeek-R1-Distill-Qwen-7B · MATH 500accuracy (%): 91.2 -> 88.6Hurt
apply TIV to R1-7B reasoningDeepSeek-R1-Distill-Qwen-7B · GSM8Kavg tokens: 725.7 -> 110.2Improved
apply TIV to R1-1.5B on out-of-domain MMLUR1-1.5B · MMLUaccuracy (%): 29.1 -> 33.7Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy